Anda di halaman 1dari 55

Contrastive Analysis, Error Analysis and Interlanguage:

Three Phases of One Goal

The mistakes or "errors" that students make in the process of learning a second or
foreign language (target language, or TL, hereafter) have always been a cause of much
concern to the teachers and textbook writers alike 1. This concern is reflected not only in
the way writers of pedagogical grammars draw attention to the potential "pitfalls" in the
TL, but also in the many lists of "common errors" prepared by experienced teachers. 2A
systematic approach to the problem of errors, in an effort to account for their linguistic
and psychological origin, regularity, predictability, variability, etc. is, however, relatively
recent. The "one goal" mentioned in the title of this chapter refers to the attempt to
facilitate the process of TL learning (and teaching) by studying the phenomenon of
"errors" within a scientific framework that is consistent with both linguistic theory and

learning theory. In so far as the three areas of research under review have this as one of
their primary goals, if not the main goal, they may be said to constitute three phases of
one goal.
The "Outreach" of the Areas of Research
This is not to imply, of course, that the areas of research mentioned in the title have this
pedagogical goal as their qnly concern. On the contrary, each of the three fields of study
has been claimed to have important contributions to make in a variety of related areas.
Contrastive Analysis (CA hereafter) is claimed to be central to all linguistic research-in
developing a general theory of language based on the discovery of the "universals" of
language, in the study of diachronic change and of dialectal variation, in longitudinal
studies of language acquisition, as well as in interlingual translation 3 (see Ferguson
1968). Error Analysis (EA, hereafter), it is claimed, is significant for the insights it
provides into the strategies employed in second language acquisition, and in turn into
the process of language learning in general (see Corder 1967). The study of
Interlanguage (IL, hereafter), it is claimed, has implications for theories of language
contact, language change and language acquisition, besides its usefulness in
describing ,special language types such as immigrant speech, non-standard dialects,
non-native varieties of language and the language of aphasics and of poetry, among
others (see Nemser 1971a; Richards 1972; Corder 1971a). Despite these many and varied
claims, it is still correct, however, to say that the primary goal of all the three areas of

research has been to facilitate TL learning by providing insights into the nature of the
learner's performance.
In addition to the diversity of claims regarding their applications, CA, EA and IL also
differ from one another in a number of respects-in their theoretical assumptions,
methodologies, the nature and scope of data considered relevant in each area, the kind
of insights they provide into the nature of TL learning, and in the implications of the
studies carried out for practical classroom teaching and materials preparation.
It is the purpose of this chapter to present a "state of the art" in each f these areas
of research from the point of view of the "one goal" explained above. In particular, with
respect to each field of study, we shall examine the current trends in theory,
methodology, claims and empirical validations thereof and its contribution to TL
teaching. The following discussion is organized in four parts-the first, second and third
parts deal with CA, EA and IL respectively and the last part is the conclusion. There will
be a good deal of overlap among the sections, but this is unavoidable given the fact that
the three fields have developed at times as rivals, and as complementary to one another
at other times.
Contrastive Analysis
Although several prominent linguists and pioneers in the field of TL pedagogy,
including Henry Sweet, Harold Palmer and Otto Jespersen, were well aware of the "pull
of the mother tongue" in learning a TL, it was Charles C. Fries who firmly established

contrastive linguistic analysis as an integral component of the methodology of TL

teaching.4 Declaring that the most effective materials (for foreign language teaching) are
based upon a scientific description of the language to be learned carefully compared
with a parallel description of the native language of the .learner (1945, p. 9),
Fries may be said to have issued the charter for modern CA. In doing so, he also made
the first move in what has turned out to be one of the most spirited controversies in the
field of foreign language teaching, namely on the role and relevance of CA, but more on
this later (see Sections 1.7 and 2.2). The challenge was taken up by Lado, whose work
Linguistics Across Cultures (1957) soon became a classic field manual for practical
contrastive studies. The Chomskyan revolution in linguistics gave a fresh impetus to CA,
not only making it possible for the comparisons to be more explicit and precise, but also
giving it what seemed to be a more solid theoretical foundation by claiming the
existence of "language universals" (but cf. Bouton 1976). The volumes of The Contrastive
Structure Series (e.g., Stockwell and Bowen 1965" Stockwell, Bowen and Martin 1965)
represent this phase of CA. The chapters from the three conferences on CA held at
Georgetown, Cambridge, and Stuttgart (Alatis 1968, Nickel 1971a, Nickel 1971b,
respectively) present scholars as, by and large, optimistic about the possibilities of CA.
But by early 1970s, CA was already open to attack on both external grounds (of
empirical validity) and internal (theoretical foundations), leading Selinker to wonder
that CA was still thriving "at a period when a serious crisis of confidence exists as to

what it is" (Selinker 1971, p. 1) and Wardhaugh to forecast a "period of quiescence" for it
(Wardhaugh 1970). CA today, however, is not entirely on the defensive-not only do
"messages of hope" keep appearing from time to time (e.g., Schachter 1974, Wode 1978),
but even the proponents of alternate approaches (EA and IL) implicitly or explicitly
incorporate CA in their methodology (see Section 3.5). If anything, the controversy
seems to have clarified the possibilities and limitations of CA and its place, along with
other components, in the task of accounting for the nature of the learner's performance.
The Rationale for CA
The rationale for undertaking contrastive studies comes mainly from three sources: (a)
practical experience of foreign language teachers; (b) studies of language contact in
bilingual situations; and (c) theory of learning.
Every experienced foreign language teacher knows that a substantial number of
persistent mistakes made by his students can be traced to the "pull of the mother
tongue." Thus, when a Hindi speaker learning English says! "The plants were all right
till we kept them in the study" in the sense of ''as long as'' or an Arabic speaker persists
in retaining a pronominal reflex of the relativized noun in his relative clauses as in "The
boy that he came" (d. Catford 1968, Schachter 1974), or to give a different type of
example, the Indian learners of English systematically replace the alveolar consonants
with their retroflex counterparts, there is no doubt that the learner is "carrying over"
patterns of the mother tongue into his TL performance. Moreover, such a "carry over"

seems to result in the largest number of deviant sentences in areas where the structures
of the native language and the TL differ the most.
Students of language contact have also noted the phenomenon of "interference,"
which Weinreich defines as "those instances of deviation from the norms of either
language which occur in the speech of bilinguals as a result of their familiarity with
more than one language" (1953, p. 1). Weinreich (1953) was the first (and perhaps still
the best) extensive study of the mechanisms of bilingual interference.
The third source that has been considered to support the CA hypothesis (see
Section 1.2) is learning theory-in particular, the theory of transfer. In its simplest form
transfer refers to the hypothesis that the learning of a task is either facilitated ("positive"
transfer) or impeded ("negative" transfer) by the previous learning- of another task,
depending on, among other things, the degree of similarity or difference obtaining
between the two tasks. The implications of transfer theory for TL learning are obvious.
(For an excellent study of the application of transfer theory to second language learning,
see Jakobovits 1969; see also Carroll 1968.)
The Contrastive Analysis Hypothesis
The "strong" version of the CA hypothesis is clearly stated by Lee (1968, p. 186). CA is
based on the assumption, he says,
1. that the prime cause, or even the sole cause, of difficulty and error in
foreign-language learning is interference coming from the learners' native

2. that the difficulties are chiefly, or wholly, due to the differences between
the two languages;
3. that the greater these differences are, the more acute the learning
difficulties will be;
4. that the results of a comparison between the two languages are needed to
predict the difficulties and errors which will occur in learning the foreign
5. that what there is to teach can best be found by comparing the two
languages and then subtracting what is common to them, so that 'what the
student has to learn equals the sum of the differences established by the
contrastive analysis.'

It must be mentioned that not all theoreticians and practitioners of CA would go

along with this version of the CA hypothesis. In particular, scholars differ on how
strongly they wish to claim for interlingual interference the pride of place among error
types, and the rather "simpliste" correlation in Lee's version, between differences in
structure and learning difficulty.5 Nevertheless, some version of this hypothesis, with

the qualifications noted above (or similar ones) is assumed by most practitioners of CA.
(For a detailed discussion of the "predictive" versus "explanatory" version of the CA
hypothesis, see Sections 1.7 and 2.2.)
CA: Its Pedagogical Claims
On the basis of these, or similar assumptions, various claims have been made as to the
potential role of CA in TL teaching. Hall (1968) asserts that the era of the uniform,
standard textbook for allieamers of a TL irrespective of their language backgrounds is
over; the structure of the textbook-selection of teaching items, degree of emphasis, kinds
of practice drills, nature of exposition, etc.--should be geared to the native language of
the learner; Nickel and Wagner (1968) also make similar claims about the crucial role of
CA in both "didactic" (limitation [selection], grading, and exposition) as well as
"methodic" (actual classroom presentation) programming (see Lado 1957 and Halliday et
al. 1964 on this point). It is also claimed (by Lado 1957) that the results of CA provide
ideal criteria for selecting testing items (for an opposite view, see Upshur 1962). It is also
generally agreed that basing teaching materials on the results of contrastive studies
necessarily entails a more "mentalistic" technique of teaching-explicit presentations of
points of contrast and similarity with the native language, involving an analytical,
cognitive activity (Rivers 1968, Jakobovits 1969, Stockwell 1968).
CA and Linguistic Models
Since comparison depends on description, there exists an inevitable implicational

relationship between CA and linguistic theory. Accordingly, the assumptions of CA, the
delicacy of its comparisons and forms of contrastive statements have changed from time
to time, reflecting the changes in linguistic theory.
The earlier contrastive studies were conducted in the structural framework, although
structural linguistics, with its insistence on describing each language in its own terms,
theoretically precluded any comparison across languages. However, characteristic of a
practice that has been endemic in CA, the theoretical and methodological contradictions
did not deter practitioners of CA. Taxonomic CA displayed the similarities and
differences between languages in terms of similarities and differences in (i) the form and
(ii) the distribution of comparable units (comparability being based on nothing more
spectacular than "gut feelings").
With the advent of generative grammar, taxonomic CA, like taxonomic
descriptive linguistics in general, has been criticized for its preoccupation with the
surface structure of language (d. Di Pietro 1968 and 1971). Three aspects of the TG
model have profoundly influenced CA: (1) the universal base hypothesis; (2) the deep
and surface structure distinction; and (3) the rigorous and explicit description of
linguistic phenomena. The universal base hypothesis, it is claimed, provides a sounder
theoretical foundation for CA as contrasted with the structuralisms relativity
hypothesis, for the assumption that all languages are alike at an abstract, underlying
level provides, theoretically at least, a basis for comparability. 6 Methodologically, the


description of this underlying level of representation in terms of a universal (nonlanguage specific) set of basic grammatical primes, semantic features and phonetic
(distinctive) features makes it easier to state similarities and differences in a uniform
manner. The explicit incorporation of two levels of linguistic organization makes it
possible for the contrastive linguist to capture and represent the intuitions of bilinguals
about the translation equivalence of utterances in two languages, although they are
disparate on the surface. Finally, the adaptation of mathematical models for the
description of natural language phenomena has enabled descriptions to be rigorous and
explicit. This in turn has enabled comparisons to be rigorous and explicit as contrasted
with, for example, statements such as

The past definite, or preterite, je portai corresponds to the English I


Modified by a fiction

The past indefinite is frequently used for the past definite in

colloquial style (cf. Halliday et al. 1964, p. 118).

This is not to say that the use of the TG model has solved the problem of CA; on


the contrary, it has made explicit the intricate problems facing CA which had not
previously been appreciated. Nevertheless, it will not be disputed that the application of
the TG model has made it possible for comparisons and contrasts to be insightful and
sophisticated to a degree unimaginable two decades ago.
CA: The Methodology
The prerequisite for any contrastive study is the availability of accurate and explicit
descriptions of the languages under comparison. It is also essential that the descriptions
be theoretically compatible. Given such descriptions, how does one go about comparing
two languages?
a. Selection
It is generally agreed that attempting to compare two languages in entirety is
both impractical and wasteful. An alternative is suggested by the British linguists, who
advocate a Firthian "polysystemic" approach. This approach is based on the assumption
that language is a "system of systems." Hence comparisons are made in terms of
particular systems and subsystems (e.g., the personal deictic system, the auxiliary verb
system, etc.) (d. Halliday et al. 1964; Catford 1968). While this ap,proach may work in
contrasting "closed systems" such as the determiner sy~tem, or even for phonology as a
whole, which can be reduced, according to one view, to an "item by item analysis of
segment types," it does not seem to be suited for syntactic comparison, which must
handle "a boundless class of possible sentences." (Cf. Langacker 1968.)


A second criterion for selection has often been advocated by scholars who
consider the role of CA to be primarily "explanatory" and not "predictive" (see Catford
1968; Lee 1968). According to these scholars, CA should limit itself to "partial"
comparisons, analyzing those parts of the grammar which are known (through error
analysis) to present the greatest difficulty to learners. But such an approach, as Hamp
rightly points out, is of limited value-we need CA to provide a "theory adequate to
explain cases not in our corpus" (1968, p. 146).
A reasonable approach to this problem is taken by Langacker (1968), who
suggests that syntactic comparison should cover approximately the same ground that
the language teacher is called on to deal with explicitly in the classroom. Within this
area, common productive processes (such as infinitive embedding, for example) should
be compared for the two languages with respect to the rules generating them.
This is essentially the approach adopted by Stockwell et al. (1965). While
Stockwell admits that their approach was somewhat tempered by [error-analysis]" (used
as a delimitation device in selection), he insists that "the most useful basis for
contrastive analysis is entirely theoretical" (1968, p. 25).
b. Comparability: the problem of equivalence
The discussion in the previous section dealt with the general problem of
selection. A much more difficult and crucial problem is that of "comparability,." i.e., of
establishing just what is to be juxtaposed for comparison. Despite the extensive study of


various aspects of CA, this problem which lies at the heart of CA has yet to be
satisfactorily, resolved. The question can be approached from three points of view, viz.,
those of (i) structural (or formal) equivalence; (ii) "translation" equivalence; (iii) both
structural and translation equivalence. While the most widely used criterion in the
literature has been that of translation equivalence, the term has been used rather loosely.
Harris seemed to work on the assumption that for a given sentence in language A there
would be only one "roughly unique" translation in language B, and proposed to
construct a "transfer" grammar on the basis of the "minimal grammatical differences"
(1954, p. 259). Levenston (1965), on the other hand, points out the possibility of multiple
translation equivalents (d. also Halliday et al. 1964, p. 121) and hence advocates

"translation paradigms"-i.e., tabulation of the various structural

configurations by which a given item may be translated, with specification of the

contextual restrictions governing the use of each equivalent. Catford, on the other hand,
believes that "the only basis for equating phonemes or for equating grammatical units in
two languages is extra-linguistic-is substantial rather than formal" (1968, p. 144). For
him, the test of translation equivalence is the interchangeability of the items in a given
situation (1965, p. 49).
Is it possible to formalize the relationship that should hold between constructions
that are considered translation equivalents by a "competent bilingual"? There have been
a few attempts to confront this crucial problem. Dingwall (1964) proposed that


"languages are more likely to be similar in their 'kernel' than in their total structure, and
that which is obligatory in the most valued grammar is more basic than that which is
optional," but with the demise of the notion of "kernel" sentences, his hypothesis has
become somewhat outdated. Perhaps the single most influential work on this question is
Krezeszowski (1971). This chapter, although it does not solve the problem of
equivalence, shows how much CA has gained in rigor and sophistication from the
application of current generative theory. In this chapter, Krezeszowski proposed that
"equivalent sentences have identical deep structures even if on the surface they are
markedly different" (1971, p. 38), "deep structures" being defined in the sense of Lakoff
(1968), in terms of basic grammatical relations, selectional restrictions and co-occurrence
relations. While this is probably the closest we have ever come to rigorously defining the
notion of "equivalence," even this formulation is still far from satisfactory, as is apparent
from the works discussed below.
Bouton (1975) points out that there are large classes of constructions which are
translation equivalents but cannot be derived from a common deep structure (in the
sense of Krezeszowski)-instances where deep structure parts contain crucial
information with regard to notions of stativity, transitivity, tense/aspect, polarity of
presupposition, etc.- thus calling for either a redefinition of "deep structure" to include
"contextual" structure or the rejection of Krezeszowski's hypothesis as it stands.
Y. Kachru (1976) has shown the limitation of a purely structural notion of


equivalence and the relevance of pragmatics and "conversational implicature" for

defining "equivalence." Fillmore (1965) had earlier pointed out instances of translation
equivalence "which are constructed along non-analogous (structural) principles" and
"cases where sentences in one language cannot be translated into another language at
all" (1965, p. 122).
A different approach to defining equivalence is suggested in Sridhar (1980). In his
cross-linguistic experimental study of sentence production, Sridhar found that common
perceptual stimuli often produced structurally different responses in different languages
which, nevertheless, were functionally similar. For example, in describing a scene in
which an inanimate object (e.g., a ball) acts upon an animate, human object (e.g., a doll),
the inherent salience of the latter causes the movement of the object NP to the sentenceinitial position, resulting in passives and topicalized sentences in English, but active
sentences with object fronting in other languages like Hungarian, Japanese, Kannada,
Turkish, etc. This technique, therefore, demonstrates the possibility of establishing
functional equivalence across structures in empirical terms.
While discussion, formalization, and refinement of the notion of equivalence
proceeds on the theoretical plane, the problems involved in this endeavor have not
significantly impeded the flow of practical contrastive studies and their application to
classroom and text materials. I will now briefly consider the state of the art in practical
contrastive analysis.


The Scope of Contrastive Studies

By "scope" here I mean the levels of linguistic structure and language use covered by
contrastive studies. Even a cursory glance at the extensive bibliographies by Hammer
and Rice (1965) and Gage (1961), as well as the volumes of IRAL, Language Learning and
other journals, reveals that the major emphasis has been on contrasting phonological
systems. Also, it is consistent with the structuralist dictum regarding the primacy of
speech. However, as Stockwell rightly reminds us, it is time to face up to the fact that
"pronunciation is simply not that important. ...Grammar and meaning are at the heart of
the matter" (1968, p. 22). Despite the "kiss of life" that syntax has received with the
advent of generative grammar, the number of sophisticated studies of contrastive syntax
still remains rather small. (Part of the problem may have to do with the rapid change in
syntactic theory in the last thirty years that has left the "applied" linguist constantly
trying to catch up with the new developments.) The best full-length studies of
contrastive syntax still remain in the volumes produced under The Contrastive Structure
Series of the Center for Applied Linguistics, Washington, D.C. The area of vocabulary
has hardly been touched at all. One of the notable exceptions is Oksaar (1972). In that
work, Oksaar reports on research using the semantic differential technique (Osgood,
Hofstatter) in order to measure intra- and interlingual differences (German-Swedish) in
the area of connotative meaning. Taking certain operational terms to demonstrate the
approach, she comes to the following conclusion: the "competing" terms differ from


each other in the two languages; and interferences are likely on the non-denotative
meaning level of the second language, the source of which lies in the influence of the
mother tongue. The extensive work done in bilingual lexicography has not been, as
Gleason correctly points out, "deeply theory-informed work" (1968, p. 40). The huge
area of usage still remains practically unchartered, and in the absence of a viable theory,
the best that can be done in this area is, in the words of Stockwell, "listing with insight."
Lado (1957) strongly advocated the need to include comparison of cultures as an
integral part of contrastive linguistics, yet his example does not seem to have been
pursued seriously. Thus the picture of contrastive studies today is rather lopsidedleaning heavily on the side of phonology, moderately inclined to syntax, but (to mix
metaphors) leaving entire flanks of lexicography, semantics and usage almost
completely exposed.
Critics of CA
For convenience of discussion, we may consider the major criticisms of CA under two
heads: (i) criticism of the predictions made by contrastive analysis and (ii) criticism of
the theoretical basis of CA.
Critics of CA have argued that since native language interference is only one of
the sources of error, indulging in CA with a view to predicting difficulties is not worth
the time spent on it; moreover, they argue, many of the difficulties predicted by CA do
not show up in the actual learner performance at all; on the other hand, many errors


that do turn up are not predicted by CA. In the light of this, they suggest, the only
version of CA that has any validity at all is the a posteriori version, i.e., the role of CA
should be explanatory, restricted to the recurrent problem areas as revealed by error
analysis, rather than the a priori or predictive version (see Whitman and Jackson 1972,
Gradman 1971, Lee 1968, Ritchie 1971, among others).
These, and some other criticisms of CA have been, in my opinion, ably answered
in James (1971). Suffice it here to say that the proponents of the strong version of CA are
the first to concede that CA does not account for all errors; they never claimed that it did
(see the caveats in Section 1.2). Secondly, the non-occurrence of error does not
necessarily invalidate the prediction-on the other hand, it may confirm it in that it
provides evidence that the student is avoiding the use of problematic structures (cf.
Corder 1973, Schachter 1964, Celce-Murcia 1978). CA cannot merely be a subcomponent
of EA because, for one thing, what we need is not only a taxonomic classification of a
corpus of data but a corpus-free theory of errors and, for another, "predictive" CA
brings to light areas of difficulty not even noticed by EA (d. Schachter 1974). Moreover,
the failure of the predictions of CA in particular instances does not necessarily
invalidate the theory itself-a distinction often lost sight of by the extremist critics of CA.
All that it shows is that we need a more precise characterization of what type of, and
under what conditions, prior linguistic knowledge is made use of (see Wode 1978 for an
attempt to define this in structural terms). After all, there have been scores of instances


in the published literature of the last decade where the predictions of CA have been
borne out by empirical results (see, for example, Duskova 1969, Schachter 1974, Bieritz
1974 among others). George (1972) estimates that approximately one-third of all errors
made by TL learners can be traced to native language interference. Therefore, as
Stockwell (1968) says, as long as one of the variables that contributes to success or
failure in language learning is the conflict between linguistic systems, CA has a place in
T1 methodology. The critics of CA have not conclusively proved this is not so. If
anything, recent developments in the theory and methodology of EA and IL have
explicitly incorporated the assumptions and methodology of CA in their models (see
Section 2.3.4). Saying that CA should be only one component among others of TL
methodology is not a criticism of CA per se-after all, it was meant to be exactly that.
Those who have attempted to "put CA in its place" may have revealed their own
The second type of criticism seeks to show that given its theoretical and
methodological assumptions, CA is in principle incapable of accounting for learner
behavior. For instance, Newmark and Reibel (1968) contend that interference is an otiose
idea and that ignorance is the real cause of errors. Dulay and Burt (1972), among others,
accuse CA of being based on the behaviorist conditioning principle, which has now
fallen on evil days. Dickerson (1974) says that CA, by denying the "variability" (i.e.,
presence of a wide assortment of pronunciations) and the "systematicity" characteristic


of the leamer's output, is necessarily forced to predict "categorical" (i.e., non-variable)

performance, which does not exist.
The argument of Newmark and Reibel (1968) has been answered by James (1971)
and I shall not go into it here. As for the second criticism, despite the authors' dismissal
of Corder's argument, I think Corder (1967) is essentially correct in claiming that CA is
not incompatible with the generativists' view of language learning as a hypothesis
testing process. Only with this view the psychological basis of "interference" would shift
to something more like that of transfer of training, in that the learner may be said to
select his experience with the learning of his native language as one of the initial
hypotheses (or "processing strategies") to be tested in the course of learning a TL. As for
the third criticism (Dickerson 1974), it must be granted that this is one of the most
serious criticisms leveled against CA and calls for a deliberate response. At this point, I
shall content myself with a few observations. There is nothing in the CA hypothesis that
denies the learner's language systematicity: in fact, the very premise of predictability is
the systematicity of the learner's performance. The presence of elements other than
those due to interlingual interference is, though correct, not a criticism of CA per se. On
the question of "variability," it is true that none of the current .models of CA
incorporates this feature. After all, variability still remains a challenge to descriptive
linguistics as well, and CA can only be as good as the description on which it is based.
Certainly variability must be accounted for in synchronic description as well as


contrastive analysis. Selinker's impressionistic observations on the emergence of

"fossilized" elements in the learner's language under certain circumstances are the first
step toward recognition and exploration of this important aspect (cf. Selinker 1972).
Krashen's (1976, 1978) "monitor" model of TL performance is another. I submit that by
treating unsuppressed or unmonitored access to native language patterns as one of the
"variables" responsible for the "variability" of TL performance, we can reconcile CA with
the variability model. Models of CA in the past have shown considerable resilience, and
claims such as that variability analysis is the "Waterloo of CA" seem to be a bit
premature at this point.
Error Analysis
Traditional EA
Of the three areas of study under review, Error Analysis (EA) has probably the longest
tradition. Yet, until recently a typical EA went little beyond impressionistic collections of
"common" errors and their taxonomic classification into categories (mistakes of
agreement, omission of articles, etc.). Little attempt was made either to define "error" in
a formally rigorous and pedagogically insightful way or to systematically account for
the occurrence of errors either in linguistic or psychological terms. Hence it is
substantially correct to say that traditional EA was an ad hoc attempt to deal with the
practical needs of the classroom teacher.


The goals of traditional EA

The goals of traditional EA were purely pragmatic-EA was conceived and

performed for its "feedback" value in designing pedagogical materials and strategies. It
was believed that EA, by identifying the areas of difficulty for the learner, could help in
(i) determining the sequence of presentation of target items in textbook and classroom,
with the difficult items following the easier ones; (ii) deciding the relative degree of
emphasis, explanation and practice required in putting across various items in the TL;
(iii) devising remedial lessons and exercises; and finally, (iv) selecting items for testing
the learner's proficiency. The "applied" emphasis in this approach to error is obvious.

c. The methodology
The methodology of EA, in so far as traditional EA can be said to have followed a
uniform method at all, consisted of the following steps:

collection of data (either from a "free" composition by students on a given

theme or from examination answers);

identification of errors (labeling, with varying degrees of precision

depending on the linguistic sophistication brought to bear on the task,
with respect to the exact nature of the deviation, e.g., dangling


preposition, anomalous sequence of tenses, etc.);


classification into error types (e.g., errors of agreement, articles, verb

forms, etc.;

statement of relative frequency of error types;

identification of the areas of difficulty in the TL; 6. therapy remedial drills,

lessons, etc.).

While the above methodology is roughly representative of the majority of error

analyses in the traditional framework, the more sophisticated investigations (for
example, Rossipal1971, Duskova 1969) went further, to include one or both of the

analysis of the source of errors (e.g., mother tongue interference,

overgeneralization, inconsistencies in the spelling system of the TL, etc.);


determination of the degree of disturbance caused by the error (or the

seriousness of the error in terms of communication, norm, etc.).

Notice that the inclusion of the two tasks just mentioned brings with it the


possibility of making EA broadbased and of evolving a theory of errors. This possibility,

however, has only recently begun to be explored.
Resurgence of Interest in EA
It was with the advent of CA and its claim to predict and explain (some major types of)
errors that serious interest began to be taken in EA. Although in the beginning CA, with
its relatively sophisticated linguistic apparatus and the strong claim to predict a
majority of errors in TL learning, seemed to condemn EA to obsolescence, as the claims
of CA came to be tested against empirical data, scholars realized that there were many
kinds of errors besides those due to interlingual interference that could neither be
predicted nor explained by CA. This led to renewed interest in the possibilities of EA.
The claim for using EA as a primary pedagogical tool was based on three arguments: (1)
EA does not suffer from the inherent limitation of CA--restriction to errors caused by
interlingual transfer: EA brings to light many other types of errors frequently made by
learners, for example, intralanguage errors arising from the particular teaching and
learning strategy employed (d. Richards 1971a). (2) EA, unlike CA, provides data on
actual, attested problems and not hypothetical problems and therefore forms a more
efficient and economical basis for designing pedagogical strategies (Lee 1968). (3) EA is
not confronted with the complex theoretical problems encountered by CA (e.g., the
problem of equivalence) (Wardhaugh 1970).
Based on arguments such as these, some scholars (e.g., Wilkins 1968) have


argued that there is no necessity for a prior comparison of grammars and that an errorbased analysis is "equally satisfactory, more fruitful, and less time consuming" (p. 102).
The experimental evidence, the little that there is, however, does not support such an
extreme position. The investigations in Duskova (1969), Banathy and Madarasz (1969),
Richards (1971b), Schachter (1974), and Celce-Murcia (1978), among others, reveal that
just as there are errors that are not handled by CA, there are those that do not surface in
EA, and that EA has its role as a testing ground for the predictions of CA as well as to
supplement its results.
The Reorientation of EA
At the same time that the extended domain of EA vis--vis CA came to be appreciated,
a development took place, largely as a result of the insights of British linguists and those
influenced by them (Corder 1967, 1971a, 1971b, 1973, 1974; Strevens 1970; Selinker 1969,
1972; Richards 1971a, 1971b, 1973) which has not only revolutionized the whole concept
of EA, but also opened up an exciting area of research commonly referred to as
Interlanguage (IL). Although in the current literature the distinction between EA and IL
is not always clear, we will, for the purpose of this chapter, study the developments in
two parts-those directly relevant to the theory and practice of EA in this part and those
having to do with IL in the next.
a On the notion of "error."
Pit Corder, in his influential paper (1967), suggested a new way of looking at the errors


made by the learner of a TL. He justified the proposed revision in viewpoint on the
basis of "the substantial similarities between the strategies employed by the infant
learning his native language and those of the second language learner." The notion of
"error," he argued, is a function of the traditional practice to take a teacher-centered
viewpoint of the learner's performance and to judge the latter in terms of the norms of
the TL. From the perspective of the language learner, the observed deviations are no
more "errors" than the first approximations of a child learning his mother tongue are
errors. Like the child struggling to acquire his language, the second-language learner is
also trying out successive hypotheses about the nature of the TL, and from this
viewpoint, the learner's "errors" (or hypotheses) are "not only inevitable but are a
necessary" part of the language learning process.
d. Errors versus mistakes
At this point, Corder introduces an important distinction between "errors" and
"mistakes." Mistakes are deviations due to performance factors such as memory
limitations (e.g., mistakes in the sequence of tenses and agreement in long sentences),
spelling pronunciations, fatigue, emotional strain, etc. They are typically random and
are readily corrected by the learner when his attention is drawn to them. Errors, on the
other hand, are systematic, consistent deviances characteristic of the learner's linguistic
system at a given stage of learning. "The key point," he asserts, is that the learner is
using a definite system of language at every point in his development, although it is not.


..that of the second language. ... The learner's errors are evidence of this system and are
themselves systematic (1967, p. 166).
Corder proposed the term "transitional competence" to refer to the intermediate
systems constructed by the learner in the process of his language learning.
e. The goals of EA
Given this redefinition of the notion of error, it follows that the goals of EA as
conceived previously also need to be redefined. In a subsequent paper (1971b), Corder
makes a distinction between the theoretical and applied goals of EA. EA has too often,
he argues, concerned itself exclusively with the "applied" goal of correcting and
eradicating the learner's errors at the expense of the more important and logically prior
task of evolving an explanatory theory of learner's performance. The study of the
systematic errors made by the learners of a TL yields valuable insights into the nature of
language learning strategies and hypotheses employed by learners and the nature of the
intermediate "functional communicative systems" or languages constructed by them.
Thus the theoretical aspect of EA is as worthy of study in and of itself as is that of child
language acquisition and can, in turn, provide insights into the process of language
acquisition in general.
f. The data and method of EA
We have already noted Corder's distinction between "mistakes" and "errors."
Corder proceeds to point out that not all errors are overtly observable, i.e., the


traditional reliance on obvious observable deviations in the learner's productive use of

the TL is not a reliable procedure for data-collection purposes. The "covertly erroneous"
utterances, i.e., utterances that are superficially well formed and acceptable, but
produced by a set of rules different from those of the TL (e.g., "1 want to know the
English" in the sense of "1 want to learn English") should also be considered part of the
data for EA. Learner's performance may also be right "by chance," i.e., due to
holophrastic learning or systematic avoidance of problem structures. All this goes to
show that the learner's errors-overt or covert-"are not properly to be regarded as right or
wrong in themselves but only as evidence of a right or wrong system" (Corder 1973, p.
274). Hence, the object of EA is to describe the whole of the learner's linguistic system
and to compare it with that of the TL. That is why error analysis is "a brand of
comparative linguistic study" (Corder 1973, p. 274).
As Corder correctly observes, the crucial element in describing the learner's
system is the correct interpretation of the learner's utterance. This is to be done, he says,
by reconstructing the correct utterance of the TL, matching the "erroneous" utterance
with its equivalent in the learner's native language. If this can be done by asking the
learner to express his intentions in the mother tongue (the translation guaranteeing its
appropriateness), then it is an authoritative reconstruction. If the learner is not available
for consultation, and the investigator has to rely on his knowledge of the learner's
system, his intentions, etc., then it can only be called plausible reconstruction (Corder


1973, p. 274).
On the basis of these data, the investigator can reconstruct the learner's linguistic
system. This is to be complemented with a psychological explanation in terms of the
learner's strategies and the process of learning. (See Section 2.3.)
Is the Notion of Error Obsolescent?
The proposed change in the attitude toward the learner's deviant utterances raises
several important questions from the pedagogical point of view. Since the assumptions
underlying the current approach to EA and IL are identical, I shall postpone discussion
of these questions until after we have examined the concept of IL in more detail. In this
section, I shall merely point out some of the issues that need to be clarified in the new
framework of EA.
For one thing, we need criteria to distinguish between productive, systematic
deviations and non-productive deviations in the learner's performance in order to make
learning more efficient. Secondly, we need criteria to determine the seriousness of
"errors" in terms of the degree of disturbance to effective communication (intelligibility,
etc.) caused by them. Thirdly, we need to reexamine the notion of errors in the context of
second-language teaching, especially in those settings where the primary object of
learning a second language is not so much to communicate with the native speakers of
that language, but for "internal" purposes, and where full-fledged, functionally
adequate non-native varieties of a TL are in wide use (for example, English in India; see


B. Kachru 1976). It is only when we have clarified these issues that EA will have a
pedagogically useful role to play. We shall take up these questions again in Section 3.6.
The successive linguistic systems that a learner constructs on his way to the mastery of a
TL have been variously referred to as "idiosyncratic dialects" (Corder 1971a),
"approximative systems" (Nemser 1971a) and "interlanguages" (Selinker 1969). The term
"Interlanguage" is becoming established in the current literature on the subject, possibly
because it is neutral as to the directionality of attitude-the other two terms imply a T1centered perspective.
The term Interlanguage (IL) seems to be appropriate also for the following
reasons: (1) it captures the indeterminate status of the learner's system between his
native language and the T1; (2) it represents the "atypical rapidity" with which the
learner's language changes, or its instability; (3) focussing on the term "language," it
explicitly recognizes the rule-governed, systematic nature of the learner's performance
and its adequacy as a functional communicative system (from the learner's point of
view, at least).
The single most important influence on the study of IL phenomena has been the
findings of the (post-structuralist) studies of child language acquisition (see Cook 1969,
1973). In a sense, the progression from traditional EA to the concept of IL may be said to
parallel the shift from the "telegraphic speech" model of child language to the recent


study of the stages of child language acquisition in sui generis terms. The earlier model
treated the child's speech as a truncated, "telegraphic" version of adult language and
proceeded to derive the child's utterances by means of deletion rules operating on the
adult system, just as EA looked upon the second-language learner's performance as
"inadequate approximations of the TL norm." Recent studies in child language
acquisition (d. Brown 1973) have recognized the absurdity of describing the child as
possessing all the rules of the adult language together with a suspiciously large number
of deletion rules. The current approach treats child language learning as a progression
of self-contained, internally structured systems, getting increasingly similar to the adult
language system. This was essentially the approach advocated as early as 1941 by
Jakobson.7 The parallelism between this change of approach in developmental
psycholinguistics and the change from traditional EA to the concept of IL is obvious.

IL: Assumptions
Defining an "approximative system" (La as a "deviant linguistic system actually
employed by the learner attempting to utilize the target language," Nemser (1971a)
states the assumptions underlying the concept of La's:

...Our assumption is three-fold: (1) Leamer's speech at a given time is the

patterned product of a linguistic system, La, distinct from LS and L


T [the source and the target language] and internally structured. (2)
La's at successive stages of learning form an evolving series Lai. ..n,
the earliest occurring when a learner first attempts to use L T, the
most advanced at the closest approach to LT ...(3) In a given contact
situation, the La's of learners at the same stage of proficiency
roughly coincide, with major variations ascribable to differences in
learning experience (p. 116).

Similar views are put forth by Corder (1967, 1971a), Selinker (1969, 1972) and
Richards (1971b, 1973). Selinker (1972) has proposed a theoretical framework to account
for IL phenomena in second language learning. According to Selinker, the most crucial
fact that any description of IL must account for is the phenomenon of fossilization.
Fossilizable linguistic phenomena are linguistic items, rules, and subsystems
which speakers of a particular NL will tend to keep in their IL relative to a particular
TL, no matter what the age of the learner or amount of explanation or instruction he
receives in the TL (po 215).
In order to account for this phenomenon, Selinker posits the existence of a
genetically determined latent psychological structure (different from Lenneberg's "latent
linguistic structure") "which is activated whenever an adult attempts to produce
meanings, which he may have, in a second language he is learning" (p. 229). This latent


psychological structure contains five central processes (language transfer, transfer of

training, strategies of second-language learning, strategies of second-language
communication and overgeneralization of TL linguistic material) and a few minor ones
(e.g., hypercorrection, spelling pronunciation, cognate pronunciation, holophrase
learning, etc.). Each process, he suggests, "forces fossilizable material upon surface IL
utterances, controlling to a very large extent the surface structures of these sentences"
(p. 217). It follows from this that "each of the analyst's predictions as to the shape of IL
utterances should be associated with one or more of these. .. processes" (p. 215).
Before proceeding to consider the suggested methodology for describing ILs in
terms of the processes listed above, it may be fruitful to clarify some of the terms used
to refer to the processes.
"Language transfer" is, of course, self-explanatory. So is "overgeneralization" (e.g.,
"What did he intended to say?"). "Transfer of training" is different from either of these,
and refers to cases such as the one where Serbo-Croatians find it hard to use the he/she
distinction in English correctly, due to the presentation of drills in textbooks and
classroom exclusively with he and never with she. An example of a "second-language
learning strategy" would be the tendency to reduce the TL to a simpler system (e.g.,
omission of function words, plural markers, etc.). "Second-language communication
strategy" would be the tendency to stop learning once the learner feels he has attained a
"functional competence" in the TL, or that certain elements in the TL are not crucial for


effective communication.
Both Selinker (1972) and Corder (1971b, 1974) agree that since ILs are internally
patterned autonomous systems, the data for IL should be based on sources other than
those used in conventional EA. Selinker argues that the only observable data from
meaningful performance in controlled situations (as opposed to classroom drills and
experiments with nonsense material) we can establish as relevant for interlingual
identification are (1) utterances in the learner's native language (NL) produced by the
learner; (2) IL utterances produced by the learner; and (3) TL utterances by the native
speakers of that TL (po 214)0
As opposed to Selinker who feels that "the analyst in the interlingual domain
cannot rely on intuitive grammatical judgments since he will gain information about
another system. ..i.e., the TL" (p. 213, 219), Corder (1974) does not consider this a
drawback, because the judgments of the learner will give crucial information on what
he thinks is the norm of the TL, thus (unconsciously) revealing his own IL system.
IL: The Empirical Evidence
Nemser (1971a) provides some arguments for the structural independence of IL from
the NL and TL systems. Based on his study of the acquisition of English phonology by
Hungarian learners (Nemser 1971b), he points out that the learner's IL (i) exhibits
frequent and systematic occurrence of elements not attributable to either the learner's
NL or TL; (ii) constitutes a system exhibiting true internal coherence when studied in


sui generis terms. Supporting his second assumption (with regard to the evolutionary
stages), Nemser notes that the amount and type of deviation in the successive stages of
language learning varies systematically, the earlier stages being characterized by
syncretism (under-differentiation), while the later stages are marked by processes of
reinterpretation, hypercorrection, etc. In order to account for the systematicity of
deviant forms (or their "fossilization" at a given stage), Nemser posits the play of two
forces: demands of communication force the establishment of phonological, grammatical
and lexical categories, and the demands of economy force the imposition of the balance
and order of the linguistic system.
Richards (1971a, 1971b) , extrapolating from the results of EA in various second
language learning situations, shows that many of the "deviant" forms produced by
learners can be accounted for in terms of one or more of the processes posited by
The acid test for the IL hypothesis would be, of course, longitudinal studies of secondlanguage learning. This task is made extremely complicated by what has earlier been
referred to as the instability of the learner's IL. In this difficult area of research, one of
the most rigorous studies to have appeared to date is Dickerson (1974).
In her study of the acquisition of selected consonant sounds of English by a
group of ]apanese learners, Dickerson demonstrates that the learner's output at every
stage is both systematic and variable, the variability being a function of the internal


linguistic environment of the sounds as well as the external style stratification (e.g., inclass versus out-of-class contexts). Dickerson's use of the Labovian variability model to
the study of TL acquisition is significant for at least two reasons: (i) methodologically, it
is ideally suited for the study of manifestly unstable language phenomena such as
learners' intermediate systems; and (ii) theoretically, it provides a more plausible
explanatory account of the so-called backsliding to IL norms noted by Selinker, Krashen,
and many others in the performance of language learners (d. Section 1.7).
It is obvious that the studies reported above seem to provide at best partial
evidence for the IL hypothesis. What the study of IL needs is empirical evidence
validating each of the psychological constructs posited by Selinker. This task is impeded
at present by the lack of a rigorous discovery procedure that can unambiguously
identify whether a given utterance in the learner's IL is produced by the operation of
one process as against another. As long as we lack such procedures, the greater
explanatory power claimed for IL will remain no more explanatory than that of the
much maligned lists of errors organized into error types.
IL in Relation to CA and EA
At this point, we may pause to consider in what respects, if any, the theory and
methodology of IL differ from those of the two other approaches to learner's
performance discussed in this chapter, and to try to assess whether this difference
actually amounts to an improvement.


The most obvious difference, of course, is in the attitude toward learner's performance,
especially toward the "errors." While CA per se does not take any position on this issue,
traditional EA considers errors to be harmful and seeks to eradicate them; in the
framework of IL, the deviations from the TL norms are treated as exponents of the
learner's system.
Secondly-and perhaps this is the most important difference-- while CA is
concerned exclusively with that aspect of the learner's performance which can be
correlated with the characteristics of his native language, IL avoids this limitation.
Native language interference is only one of the explanatory tools in the repertoire of the
IL investigator. Thus IL is explanatorily more powerful in as much as it includes the
explanatory power of CA and extends beyond it.
Methodologically, IL may be said to incorporate the assumptions of both CA and
EA. While CA contrasts the learner's native language and the TL, and conventional EA
involves contrast between the learner's performance and the TL, IL takes all three
systems into account, explicitly incorporating the contrastive analysis of the learner's IL
with both his native language and the TL. The difference is that, in IL, the contrastive
analysis is an initial filtering device, making way for the testing of hypotheses about the
other determinants of the learner's language.
Pedagogical Implications
It is perhaps too early to expect concrete suggestions for practical classroom teaching


and preparation of materials based on the assumptions of the new approach to EA and
the study of IL. Yet one may wish to at least speculate on the possible pedagogical
implications of the recent studies, if only to generate controversy.
A major outcome of the application of IL studies to TL pedagogy would be a
radical change in the teacher's attitude toward the learner's performance (cf. Corder
1967, Cook 1969, Richards 1971a). In particular, the teacher should give up the
unreasonable expectation of TL performance from the learner from the very start.
Instead, as Dickerson (1974) suggests, he is asked to "expect variability," to measure the
learner's attempts not in terms of the TL versus non- TL opposition, but in terms of the
"proportion of TL and non- TL variants" in the learner's performance in a given
stylistic/sociolinguistic situation. From this it follows that the so-called backsliding to
the IL norm does not indicate regression but a natural sensitivity to style difference.
This in turn suggests that the traditional monolithic format of proficiency tests should
give way to a more flexible, multi-factor format sampling learner performance in
various styles and structural environments.
A similar change in viewpoint is also warranted in deciding on the model of
instruction (and consequently the norms of correctness) in those second language (or
dialect) learning contexts where indigenous non-native varieties of a TL or "nonstandard" native varieties of the TL are in wide use. English in India and West Africa,
and Black English in the U.S. are cases in point. While the systematicity, contextual


determination and functional adequacy of these varieties have been recognized for
some time now (see B. Kachru 1966, Labov 1969, among others), the pedagogical
problems posed by them are only recently being appreciated. Richards (1972) suggests
that these varieties are properly to be regarded as ILs which have developed as a result
of the particular social contexts of their learning and use. In these contexts, he suggests,
we need to distinguish not only between "errors" versus "non-errors" but also between
"errors" and "deviations," in the sense of Kachru (1966). According to Kachru,
"deviations" are explainable in terms of the sociocultural context in which English
functions in India, while "errors" are breaches of the linguistic code of English. Richards
(1972) points out the relevance of this distinction in second/foreign language teaching:
"in the foreign language setting," he observes, "all differences between the learner's use
of English and overseas English are mistakes (= errors) or signs of incomplete learning"
(p. 182); there is no scope for "deviations" here, whereas in a second-language setting
deviations (in the sense just defined) are "reflections of interlingual creativity" (p. 181).
Given this distinction, it follows that questions of instructional model, etc. are to be
decided keeping in mind the pragmatics of language use in such contexts (d. B. Kachru
1976). If the TL is to be used primarily for communicating with other members of the
interlingual community and only very marginally for communicating with the native
speakers of the TL, one wonders if the enormous time, effort and resources expended
on polishing the t's, d's, 's and s of the learners is justified. Thus the notion of "error"


in such learning contexts needs a redefinition.

This does not mean that teachers are asked to abandon comparison of the
learner's language with the norms of the TL altogether and replace the notion of error
with that of interlanguage. On the contrary, as Zyatiss (1974) remarks, a pedagogically
oriented description of the learner's language is "always contrastive and eventually
evaluative" (p. 234). This viewpoint is shared by Richards (1971b), who agrees that we
still need the notion of "errors,'~ and to "correct " them simply because speech is linked
to attitudes and social structure. Deviancy from grammatical or phonological norms of
a speech community elicits evaluational reactions that may classify a person
unfavorably (p. 21).
To sum up some of the problems raised in this section and in Section 2.4, if the
proposed reorientation in the perspective toward learner's errors is to be pedagogically
useful, we need to clarify the following: (i) the criteria to distinguish between errors
which are productive hypotheses and errors resulting from false generalizations; (ii) the
methodology to clearly identify the sources of errors in terms of the processes outlined
in Section 3.2; (iii) a hierarchy of types of errors in terms of their disturbance to effective
communication and attitudinal reactions; and finally, (iv) the notion of "error" versus
acceptable "deviations" in second-language learning contexts.
In the course of this chapter, I have attempted to show that CA, EA and IL may be


looked upon as three evolutionary phases in the attempt to understand and explain the
nature of the TL learner's performance. This "evolution" may be said to involve an
extension of perspective in many ways-in the attitude toward the learner's "errors," in
the explanatory hypotheses regarding the source(s) of the "errors," in the data
considered relevant for study and in the suggested methodology. In other words, the
approach toward the learner's performance has become more broadbased in trying to
come up with an explanatory account of why the TL learner speaks and writes the way
he does.
Perhaps the most significant outcome of the research into TL acquisition over the
last decade or so is the recognition of the similarities with first language acquisition,
both in strategies (e.g., overgeneralization) and developmental sequences (e.g., in the
order of acquisition of grammatical morphemes, d. Brown 1973). Thus TL acquisition is
viewed as a process of "creative construction" (Dulay and Burt 1976). However, this
recognition of creativity is somehow felt to be inconsistent with the notion of
interference. Thus one sees a tendency in the current literature to downplay the role of
first language interference, and an overeagerness to explain away what seem to be
patently interference errors in terms of some other strategy felt to be more respectable
or consistent with the view of the TL learner as an active experimenter with language.
Yet, as I have tried to show earlier, creativity and transfer are not at all
incompatible with each other, and to suggest that transfer theory necessarily


presupposes a "conditioning" model of learning betrays a naive understanding of both

transfer and creativity. After all, child language acquisition research is full of instances
of transfer-for, what else is "overgeneralization," the single most important strategy in
language learning, if not the transfer of hypotheses formed on the basis of previous
experience to new situations? The TL learner's errors arising from first language
interference are no more instances of conditioning than are the child's overregularizations. If the latter can be touted as instances of creative construction, as
indeed they have been, there seems to be no reason why the same explanation could not
be given of the former.
What this implies, therefore, is not that the CA approach should be thrown
overboard, but that more rigorous research is needed to identify the precise conditions
under which the TL learner utilizes the hypotheses developed on the basis of his
experience with the first language. Consequently, while one readily grants that an
explanatory account of TL learner's performance must include other components
besides interlingual interference, CA still remains the most rigorously worked-out
component of the theory. The next few years will probably see a flurry of proposals for
the study of the other major processes claimed to influence the TL learner's


Alans James E., ed. Contrastive Linguistics and Its Pedagogical Implications.
Monograph Series on Languages and Linguistics, no. 21. Washington, D.C.:
Georgetown University, 1968.
Banathy, B.H. and Madarasz, P.H. "Contrastive Analysis and Error Analysis." Journal of
English As a Second Language 4 (1969): 77-92.
Bieritz, W.D. "Phonologische Interferenzen-Franzosiche Aussprachefehler Deutscher
Studenten. International Review of Applied Linguistics 12 (1974): 193-230.
Bouton, Lawrence F. "The Problem of Equivalence in Contrastive Analysis." International
Review of Applied Linguistics 14 (1976): 143-163.
Brown, Roger. A First Language: The Early Stages. Cambridge, Mass.: Harvard University
Press, 1973.
Carroll, John B. "Contrastive Linguistics and Interference Theory." In Alatis (1968): 113122.
Catford, J.C. A Linguistic Theory of Translation. London: Oxford University Press,
1965. -."Contrastive Analysis and Language Teaching." In Alatis (1968): 159173.
Celce-Murcia, Marianne. "The Simultaneous Acquisition of English and French in a
Two- Year-Old Child." In Hatch (1978) pp. 38-53.
Cook, Vivian J. "The Analogy Between First and Second Language Learning."


International Review of Applied Linguistics 7 (1969): 207-216. -."The Comparison of

Language Development in Native Children and Foreign Adults." International
Review of Applied Linguistics 11 (1973): 13-28.
Corder, Stephen PIT. "The Significance of Learner's Errors." International Review of
Applied Linguistics 5 (1967): 161-170. -."Idiosyncratic Dialects and En-or Analysis."
International Review of Applied Linguistics 9 (1971a): 147-159. -"Describing the
Language Learner's Language." CILT Reports and Papers 6 (1971b): 57-64.
-.Introducing Applied Linguistics. Hamondsworth: Penguin, 1973. -."The
Elicitation of Interlanguage." International Review of Applied Linguistics (Special
Issue on the Occasion of B. Malmberg's 60th Birthday, edited by G. Nickel) (1974):
Dickerson, Lonna J. "Internal and External Patterning of Phonological Variability in the
Speech of Japanese Learners of English: Toward a Theory of Second Language
Acquisition." Ph.D. dissertation. Urbana-Champaign: University of Illinois, 1974.
Dingwall, William O. "Transformational Generative Grammar and Contrastive
Analysis." Language Learning 14 (1964): 147-160.
Di Pietro, R. "Contrastive Analysis and Notions of Deep and Surface Gram- mar." In
Alatis (1968): 65-80. -.Language Structures in Contrast. Rowley, Mass.: Newbury
House, 1971. DULAY, HEIDI C. and BURT, MARINA K. "Goofing: An Indicator
of Children's Second Language Learning Strategies." Language Learning 22 (1972):


235-252. -."Creative Construction in Second Language Teaching and Learning." In

Papers in Second Language Acquisition, edited by H.D. Brown. Ann Arbor, Mich.:
University of Michigan Press, 1976.
Duskov A, L. "On the Sources of Errors in Foreign Language Learning." International
Review of Applied Linguistics 7 (1969): 11-30.
Ferguson, Charles A. "Contrastive Analysis and Language Development." In Alatis
(1968) pp. 101-112.
Fillmore, C.J. "On the Notion of 'Equivalent Sentence Structure.' "The Ohio State
University Project on Linguistic Analysis. Report 11, 1965.
Fries, Charles C. Teaching and Learning English as a Foreign Language. Ann Arbor:
University of Michigan Press, 1945.
Gage, William W. Contrastive Studies in Linguistics. A Bibliographical Checklist.
Washington, D.C.: Center for Applied Linguistics, 1961.
George, H.V. Common Errors in Language Learning. Rowley, Mass.: Newbury House, 1972.
Gleason, H.A., JR. "Contrastive Analysis and Discourse Structure." In Alatis (1968) pp.
Gradman, J. "The Limitations of Contrastive Analysis Predictions." Working Papers in
Linguistics, University of Hawaii 3, no. 4 (1971): 73-77.
Hall, Robert A., JR. "Contrastive Grammar and Textbook Structure." In Alatis (1968) pp.


Halliday, Michael A.K.; McIntosh, A.; and Strevens, P. The Linguistic Sciences and
Language Teaching. London: Longmans, 1964.
Hammer, John J. (in consultation with Frank A. Rice). A Bibliography of Contras- tive
Linguistics. Washington, D.C.: Center for Applied Linguistics, 1965.
Hamp, Eric P. "What a Contrastive Grammar Is Not, If It Is." In Alatis (1968) pp. 137-147.
Harris, Zellig S. "Transfer Grammar." International Journal of American Linguistics 20
(1954): 259-270.
Hatch, Evelyn M. Set:ond Language Acquisition. Rowley, Mass.: Newbury House, 1978.
Jackson, Kenneth L. "The Verification and Comparison of Contrastive Analyses."
Working Papers in Linguistics, University of Hawaii; 3, no. 4 (1971): 201-210.
Jakobovits, Leon A. "Second Language Learning and Transfer Theory." Language
Learning 19 (1969): 55-86.
Jakobson, Roman. Kindersprache, Aphasie und Allgemeine Lautgesetze. Uppsala, 1941.
English version: Child Language, Aphasia, and Linguistic Theory. The Hague:
Mouton, 1968.
James, Carl. "The Exculpation of Contrastive Linguistics." In Nickel (1971a) pp. I 53-68.
Jespersen, Otto. How to Teach a Foreign Language. New York: Maclnillan, 1904.
Kachru, Braj B. "Indian English: A Study in Contextualization." InIn Memory of J.R. Firth,
edited by Bazell, C.C. et al. London: Longmans, 1966. -."Models of English for the
Third World: White Man's Linguistic Burden or Language Pragmatics?" TESOL


Quarterly 10 (1976): 221-239.

Kachru, Yamuna. "Defining 'Equivalence' in Contrastive Analysis; Causative
Constructions in South Asian Languages." The Bulletin of the Central Institute of
English and Foreign Languages 12 (1976): 1-13.
Krashen, Stephen. "Monitor Theory: A Model of Adult Second Language Performance."
Paper presented at the USC-UCLA Second Language Acquisition Forum, 1976.
-.'1ndividual Variation in the Use of the Monitor." In Second Language Acquisition
Research: Issues and Implications, edited by William C. Ritchie. New York:
Academic Press, 1978.
Krezeszowski, Tomasz P. Equivalence, Congruence and Deep Structure. In Nickel
(1971a) pp. 37-48.
Labov, William. "The Logic of Non-Standard English." Monograph Series on Languages
and Linguistics, no. 22. Washington, D.C.: Georgetown University, 1969.
Lado, Robert. Linguistics across Cultures. Ann Arbor: University of Michigan Press,
Lakoff, George. "Instrumental Adverbs and the Concept of Deep Structure." Foundations
of Language 4 (1968): 4-27.
Langacker, Ronald W. "Review of Stockwell et al., 1965 and Stockwell and Bowen, 1965."
Foundations of Language 4 (1968): 211-218.
Lee, W.R. "Thoughts on Contrastive Linguistics in the Context of Foreign Language


Teaching." In Alatis (1968) pp. 185-194. .

Levenston, E.A. "The 'Translation Paradigm': A Technique for Contrastive Syntax."
International Review of Applied Linguistics 3 (1965): 223-228.
Nemser, William. "Approximative Systems of Foreign Language Learners." International
Review of Applied Linguistics 9 (1971a): 115-124. -."An Experimental Study of
Phonological Interference in the English of Hungarians." (Revision of his 1961
dissertation). Bloomington: Indiana University Press, 1971b.
Newmark, L. and Reibel, D. "Necessity and Sufficiency in Language Learning."
International Review of Applied Linguistics 6 (1968): 145-164.
Nickel, Gerhard, ed. Papers in Contrastive Linguistics. Cambridge: Cambridge University
Press, 1971a. -, ed. Papers from the International Symposium on Applied
Contrastive Linguistics. Stuttgart, Bielfield: Cornelsen-Vellagen and Klasing,
Nickel, Gerhard and Wagner, K.H. "Contrastive Linguistics and Language Teaching."
International Review of Applied Linguistics 6 (1968): 233-255.
Oksaar, E. "On Contrastive Semantics: Experiments with the Polarity-Profile Method." In
Nickel (1971b).
Palmer, Harold E. The Principles of Language Study. London: Oxford University Press,
Richards, Jack C. "A Non-Contrastive Approach to Error Analysis." English Language


Teaching 25 (1971a): 204-219. -."Error Analysis and Second Language Strategies."

Language Sciences 17 (1971b): 12-22. -."Social Factors, Interlanguage, and Language
Learning." Language Learning 17 (1972): 159-188.
Ritchie, W. C. "Some Implications of Generative Grammar for the Construction of
Courses in English as a Foreign Language." Language Learning 17 (1967): 45131.
Rivers, Wilga M. "Contrastive Linguistics in Textbook and Classroom." In Alatis (1968)
pp. 151-154.
Rossipal, Ham felanalys och om en felanalys av svenskars tysk. Sprakforskning i
relation till sprekundervisning. (Edited by B. Lindgren et al.). Copenhagen.
Nordik Sommer-universitets Generalsekretariat, Fiolstraede 26. 1972, 109-127.
Schachter, Jacquelyn. "An Error in Error Analysis." Language Learning 24 (1974):
Selinker, Larry F. "Language Transfer." General linguistics 9 (1969): 67-92. -."A Brief
Reappraisal of Contrastive Linguistics." Working Papers in Linguistics, University
of Hawaii (1971): 1-10. -."Interlanguage." International Review of Applied
Linguistics 10 (1972): 209-231.
Sridhar, S.N. Perceptual and Cognitive Determinants of Syntactic Structures: A CrossLinguistic Study in Sentence Production. Urbana-Champaign: University of
Illinois Ph.D. dissertation, 1980.


Stockwell, Robert P. "Contrastive Linguistics and Lapsed Time." In Alatis (1968) pp. 1126.
Stockwell, Robert P. and Bowen, J.D. The Sounds of English and Spanish. Chicago: The
University of Chicago Press, 1965.
Stockwell, Robert P., Bowen, J.D. and Martin, J.W. The Grammatical Structures of English
and Spanish. Chicago: The University of Chicago Press, 1965.
Strevens, Peter S. "Two Ways of Looking at Error Analysis." ERIC: ED 037 714, 1970.
Sweet, Henry. The Practical Study of Languages. London: J.M. Dent, 1899. (Reprinted,
London: Oxford University Press, 1964).
Upshur, John A. "Language Proficiency Testing and the Contrastive Analysis Dilemma."
Language Learning 12 (1962): 123-127.
Wardhaugh, Ronald. "The Contrastive Analysis Hypothesis." TESOL Quartlerly 4 (1970):
Weinreich, Uriel. Languages in Contact: Findings and Problems. New York: Linguistic Circle
of New York, 1953.
Whitman, R.A. "Contrastive Analysis: Problems and Procedures." Language Learning 20
(1970): 191-197.
Whitman, R.A. and Jackson, K.L. "The Unpredictability of Contrastive Analysis."
Language Learning 22 (1972): 29-41.
Wilkins, D.A. "Review of Trends in Language Teaching edited by Valdman."


International Review of Applied Linguistics 6 (1968): 99-107.

Wode, Henning. "Developmental Sequences in Naturalistic L2 Acquisition." In Hatch
(1978), pp. 101-117.
Zyatiss, Woflgang. "A 'Kiss of Life' for the Notion of Error." International Review of
Applied Linguistics 12 (1974): 231-238.



I am grateful to Professors Braj Kachru and Yamuna Kachru for their suggestions on an

earlier version of this chapter.

See, for example, the following: George Whitworth, Indian English: An Examination of the

Errors of Idiom Made by Indians in Writing English (Letch- worth: Herts, 1907); T.L.M. PearseSmith, English Errors in Indian Schools (London: Oxford University Press, 1934); F.Q. French,
Common Errors in English (London: Oxford University Press, 1949).

The possibility of evolving a scientific theory of translation that could, in turn, be used in

machine translation has been one of the additional motivations for pursuing CA (see Catford

Q. Sweet (1899): "There is another class of difficulties which may be regarded as

partly external, partly internal-those which depend on the relations of the foreign language to
the learner's native language, especially as regards similarity in vocabulary and structure"
(pp. 53-54 in the 1964 edition). Sweet warned against the formation of wrong "crossassociations" across seemingly similar items in "closely allied languages." Jespersen
recognized NL interference, but advocated comparative analysis only as an "interesting"
adjunct to the main task of teaching the TL. "Comparisons between the languages which the
pupils know, for the purpose of showing their differences of economy in the use of linguistic
means of expression ...may often become very interesting, especially for advanced students

The teacher may call attention to the inconsistency of the languages; what is distinctly
expressed in one case is in another case not designated by any outward sign (haus: hauser'.
..sheep: sheep)" Oespersen 1904, p. 135). H.E. Palmer deals at some length with the
"illegitimate" substitutions made by English learners in speaking French-in phonology, lexis
and grammar. He also recognizes cases of positive transfer. However, he sternly warns against
"the temptation to replace habit-forming by analysis and synthesis of problem items" (Palmer
1964, p. 58).

This view seems to derive from Lado (1957, p. 2): "Those elements that are similar to his

native language will be simpler for him, and those elements that are different will be

Bouton (1976) points out that the universal base hypothesis and the notion of equivalence in

the sense of Krezeszowski are not strictly compatible.

7. See Jakobson (1941). In the words of Ferguson (1968), ". ..Jakobson made clear the notion
that a child's language is always a coherent system [al- though with more marginal features
and fluctuation than adult language] and that the development of a child's language may
profitably be regarded as a succession of stages, just as the history of a language may be."

See Jakobson (1941). In the words of Ferguson (1968), ". ..Jakobson made clear the notion
that a child's language is always a coherent system [al- though with more marginal features
and fluctuation than adult language] and that the development of a child's language may
profitably be regarded as a succession of stages, just as the history of a language may be."